For food security issues or global climate change, there is a growing need for large-scale knowledge of soil organic carbon (SOC) contents in agricultural soils. To capture and quantify SOC contents at a field scale, Earth Observation (EO) can be a valuable data source for area-wide mapping. The extraction of exposed soils from EO data is challenging due to temporal or permanent vegetation cover, the influence of soil moisture or the condition of the soil surface. Compositing techniques of multitemporal satellite images provide an alternative to retrieve exposed soils and to produce a data source. The repeatable soil composites, containing averaged exposed soil areas over several years, are relatively independent from seasonal soil moisture and surface conditions and provide a new EO-based data source that can be used to estimate SOC contents over large geographical areas with a high spatial resolution. Here, we applied the Soil Composite Mapping Processor (SCMaP) to the Landsat archive between 1984 and 2014 of images covering Bavaria, Germany. Compared to existing SOC modeling approaches based on single scenes, the 30-year SCMaP soil reflectance composite (SRC) with a spatial resolution of 30 m is used. The SRC spectral information is correlated with point soil data using different machine learning algorithms to estimate the SOC contents in cropland topsoils of Bavaria. We developed a pre-processing technique to address the issue of combining point information with EO pixels for the purpose of modeling. We applied different modeling methods often used in EO soil studies to choose the best SOC prediction model. Based on the model accuracies and performances, the Random Forest (RF) showed the best capabilities to predict the SOC contents in Bavaria (R² = 0.67, RMSE = 1.24%, RPD = 1.77, CCC = 0.78). We further validated the model results with an independent dataset. The comparison between the measured and predicted SOC contents showed a mean difference of 0.11% SOC using the best RF model. The SCMaP SRC is a promising approach to predict the spatial SOC distribution over large geographical extents with a high spatial resolution (30 m).

In many real world settings, imbalanced data impedes model performance of learning algorithms, like neural networks, mostly for rare cases. This is especially problematic for tasks focusing on these rare occurrences. For example, when estimating precipitation, extreme rainfall events are scarce but important considering their potential consequences. While there are numerous well studied solutions for classification settings, most of them cannot be applied to regression easily. Of the few solutions for regression tasks, barely any have explored cost-sensitive learning which is known to have advantages compared to sampling-based methods in classification tasks. In this work, we propose a sample weighting approach for imbalanced regression datasets called DenseWeight and a cost-sensitive learning approach for neural network regression with imbalanced data called DenseLoss based on our weighting scheme. DenseWeight weights data points according to their target value rarities through kernel density estimation (KDE). DenseLoss adjusts each data point’s influence on the loss according to DenseWeight, giving rare data points more influence on model training compared to common data points. We show on multiple differently distributed datasets that DenseLoss significantly improves model performance for rare data points through its density-based weighting scheme. Additionally, we compare DenseLoss to the state-of-the-art method SMOGN, finding that our method mostly yields better performance. Our approach provides more control over model training as it enables us to actively decide on the trade-off between focusing on common or rare cases through a single hyperparameter, allowing the training of better models for rare data points.

Precision beekeeping allows to monitor bees' living conditions by equipping beehives with sensors. The data recorded by these hives can be analyzed by machine learning models to learn behavioral patterns of or search for unusual events in bee colonies. One typical target is the early detection of bee swarming as apiarists want to avoid this due to economical reasons. Advanced methods should be able to detect any other unusual or abnormal behavior arising from illness of bees or from technical reasons, e.g. sensor failure.In this position paper we present an autoencoder, a deep learning model, which detects any type of anomaly in data independent of its origin. Our model is able to reveal the same swarms as a simple rule-based swarm detection algorithm but is also triggered by any other anomaly. We evaluated our model on real world data sets that were collected on different hives and with different sensor setups.

No abstract

Recent studies have shown that many deep metric learning loss functions perform very similarly under the same experimental conditions. One potential reason for this unexpected result is that all losses let the network focus on similar image regions or properties. In this paper, we investigate this by conducting a two-step analysis to extract and compare the learned visual features of the same model architecture trained with different loss functions: First, we compare the learned features on the pixel level by correlating saliency maps of the same input images. Second, we compare the clustering of embeddings for several image properties, e.g. object color or illumination. To provide independent control over these properties, photo-realistic 3D car renders similar to images in the Cars196 dataset are generated. In our analysis, we compare 14 pretrained models from a recent study and find that, even though all models perform similarly, different loss functions can guide the model to learn different features. We especially find differences between classification and ranking based losses. Our analysis also shows that some seemingly irrelevant properties can have significant influence on the resulting embedding. We encourage researchers from the deep metric learning community to use our methods to get insights into the features learned by their proposed methods.